def evaluate(candidate: CandidateFeature, classifier, grid_search_parameters, preprocessed_folds, score, train_data, current_target, train_X_all, train_y_all_target, test_data, test_target):
pipeline = Pipeline([('features', FeatureUnion(
[
(candidate.get_name(), candidate.pipeline)
])),
('classifier', classifier())
])
refit = False
if Config.get_default('score.test', 'False') == 'True' and not Config.get_default('instance.selection',
'False') == 'True':
refit = True
clf = GridSearchCV(pipeline, grid_search_parameters, cv=preprocessed_folds, scoring=score, iid=False,
error_score='raise', refit=refit)
clf.fit(train_data, current_target) #dataset.splitted_values['train']
candidate.runtime_properties['score'] = clf.best_score_
candidate.runtime_properties['hyperparameters'] = clf.best_params_
if Config.get_default('score.test', 'False') == 'True':
if Config.get_default('instance.selection', 'False') == 'True':
clf = GridSearchCV(pipeline, grid_search_parameters, cv=preprocessed_folds, scoring=score,
iid=False, error_score='raise', refit=True)
clf.fit(train_X_all, train_y_all_target)
candidate.runtime_properties['test_score'] = clf.score(test_data, test_target) #self.dataset.splitted_values['test']
else:
candidate.runtime_properties['test_score'] = 0.0
return candidate
def evaluate(candidate: CandidateFeature,
classifier,
grid_search_parameters,
preprocessed_folds,
score,
train_data,
current_target,
train_X_all,
train_y_all_target,
test_data,
test_target,
cv_jobs=1):
pipeline = Pipeline([('features',
FeatureUnion([(candidate.get_name(),
candidate.pipeline)])),
('classifier', classifier())])
refit = False
if Config.get_default('score.test',
'False') == 'True' and not Config.get_default(
'instance.selection', 'False') == 'True':
refit = True
print(grid_search_parameters)
clf = GridSearchCV(pipeline,
grid_search_parameters,
cv=preprocessed_folds,
scoring=score,
iid=False,
error_score='raise',
refit=refit,
n_jobs=cv_jobs)
clf.fit(train_data, current_target) #dataset.splitted_values['train']
candidate.runtime_properties['score'] = clf.best_score_
candidate.runtime_properties['hyperparameters'] = clf.best_params_
#for
test_fold_predictions = []
for fold in range(len(preprocessed_folds)):
test_fold_predictions.append(
clf.predict(train_data[preprocessed_folds[fold][1]]) ==
current_target[preprocessed_folds[fold][1]])
candidate.runtime_properties[
'test_fold_predictions'] = test_fold_predictions
if Config.get_default('score.test',
'False') == 'True' and len(test_data) > 0:
if Config.get_default('instance.selection', 'False') == 'True':
clf = GridSearchCV(pipeline,
grid_search_parameters,
cv=preprocessed_folds,
scoring=score,
iid=False,
error_score='raise',
refit=True)
clf.fit(train_X_all, train_y_all_target)
candidate.runtime_properties['test_score'] = clf.score(
test_data, test_target) #self.dataset.splitted_values['test']
else:
candidate.runtime_properties['test_score'] = 0.0
return candidate